Upload README.md with huggingface_hub

README.md

@@ -34,6 +34,20 @@ underneath.
 
 47.83 B parameters, 2 361 tensors, `cortiq verify` clean.
 
+## What comes out
+
+<video controls muted playsinline width="512"
+  src="https://huggingface.co/infosave/MiniMax-H3-Turbo-cmf/resolve/main/samples/corgi_512x288_4step.mp4">
+</video>
+
+*"A corgi in a chef hat flipping a pancake, sizzling sounds and a cheerful bark."*
+— 512×288, 39 frames at 24 fps, seed 42, **four steps**, nothing but the prompt.
+The audio is not a second model: the same transformer denoises both streams in
+one packed sequence. Also in the repository as
+[`samples/`](https://huggingface.co/infosave/MiniMax-H3-Turbo-cmf/tree/main/samples):
+the AVI `cortiq animate` actually wrote, its `.wav`, and this mp4 (remuxed for
+the browser — the runtime itself never touches ffmpeg).
+
 The LoRA is not a separate download: it is merged into the weights, so the file
 IS the 4-step model.